Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · NMA in Exercise Research

The Use of Network Meta-Analysis in Physical Exercise and Health Promotion Research

A 2024 BMJ network meta-analysis of 218 trials and 14,170 participants ranked specific exercise types against one another for depression, a comparison that a standard pairwise meta-analysis of exercise versus no exercise cannot make. The best-supported modality was not the one most commonly prescribed.

Pubrica AcademyWritten by an exercise science research methodologist Reviewed byan evidence synthesis statistician Published10 October 2026 Last checked10 October 2026 Reading time9 minutes Sources10 peer-reviewed

Network meta-analysis (NMA) combines direct evidence from trials that compared two interventions head-to-head with indirect evidence from trials that each compared a different intervention against a shared common comparator, producing a single ranked comparison across all interventions in the evidence base simultaneously1. This property is particularly important in physical exercise and health promotion research, in which trials rarely compare, for example, yoga directly against strength training but frequently compare each against a waitlist or usual-care control. A standard pairwise meta-analysis can establish only that exercise is superior to no exercise, whereas an NMA can estimate how specific exercise types compare with one another even when no trial has compared them directly. This capability is especially valuable in a field in which each intervention is supported by a relatively modest body of dedicated trials: individually, a meta-analysis of resistance training for depression or a separate meta-analysis of yoga for depression might be too small to produce a confident estimate, but within a single connected network every trial contributes to every comparison through its shared comparators. This also explains why network meta-analysis has become the preferred evidence-synthesis method behind many exercise-prescription guidelines issued by professional bodies, which must compare a full range of options against one another rather than endorse only the intervention type with the largest dedicated trial literature.

Learning objectives

  • The estimates that an NMA can provide in exercise research and that a pairwise meta-analysis cannot
  • The rankings and effect sizes produced by a large NMA of exercise types for depression
  • The reasons exercise-specific trial features (dose, adherence and blinding) create particular challenges for NMA in this field

Common misconceptions

  • “All forms of exercise are roughly equivalent for a given health outcome once you account for supervision.” The largest network meta-analysis of exercise for depression, covering 218 trials and 14,170 participants, found effect sizes that varied by more than two-fold across exercise types compared with active controls, from dance (Hedges’ g -0.96) and walking or jogging (-0.62 to -0.63) at one end to tai chi or qigong (-0.42) at the other; a spread of this magnitude precludes treating the modalities as interchangeable.

Suitability of network meta-analysis (NMA) for exercise and health-promotion evidence

Exercise and health-promotion trials typically compare one intervention arm against a passive or usual-care control rather than against every other plausible intervention type, so that most pairwise comparisons between active interventions have not been studied directly2. Network meta-analysis addresses this limitation by connecting every trial into a single network through shared comparators, allowing indirect evidence to inform comparisons that no trial tested directly, provided that the connected trials are sufficiently similar in their populations and designs for the comparison to be valid, a requirement known as transitivity34. For exercise outcomes specifically, this requires verifying that baseline fitness, age range, condition severity and level of supervision are reasonably comparable across every trial that contributes to a given shared comparator, since any of these factors can plausibly modify the benefit that a given exercise modality produces. The ISPOR Task Force’s good-practice guidance on indirect treatment comparisons describes this approach as valuable specifically in fields in which head-to-head trials of every pair of interventions would be impractical, a description that applies well to exercise research, since funding rarely supports trials comparing, for instance, resistance training directly against tai chi5. This structural feature of exercise trial design is not a temporary limitation of the current evidence base; it reflects the fact that exercise and health-promotion trials are usually funded and designed to answer a specific, local question, namely whether a given programme is more effective than no programme, rather than to construct a complete comparative map of every intervention a field has studied. Network meta-analysis enables researchers to derive such a comparative map from an evidence base that was not designed to produce one.

Ranking of exercise types for depression in a 218-trial NMA

Noetel and colleagues published a network meta-analysis in the BMJ in 2024 that synthesized 218 randomized controlled trials and 14,170 participants to compare the effect of different exercise modalities on depression6. Against active controls, the largest effects were found for dance (Hedges' g -0.96, though based on only five comparisons and 107 participants) and walking or jogging (-0.62 to -0.63), followed by yoga (-0.55), strength training (-0.49), mixed aerobic exercise (-0.43), and tai chi or qigong (-0.42)6. Strength training and yoga also showed better acceptability, with lower odds of dropout than active controls (odds ratios of 0.55 and 0.57, respectively)6. The authors applied the CINeMA framework to rate certainty in each comparison and found that only the walking-or-jogging estimate reached low certainty; every other modality's evidence was rated very low, and only one of the 218 trials met the Cochrane criteria for low risk of bias overall, mainly because blinding participants and staff to an exercise assignment is rarely possible67. The scale of this network, with 218 trials contributing to a single connected model, partly explains why the findings carry weight: a network of this size provides the statistical model with many more shared comparators through which to borrow strength for any individual comparison than several smaller, separately conducted meta-analyses could provide, which partly explains why the authors were able to report usable estimates even for modalities, such as dance, supported by comparatively few trials.

Ranking of intervention designs for increasing physical activity: a second case

A separate 2025 network meta-analysis by Hodkinson and colleagues in BMC Medicine adopted a different approach, ranking intervention designs rather than exercise types; it compared education-based, behaviour-change, motivational/goal-setting and multi-component interventions for increasing physical activity in adults with cardiometabolic conditions8. Across 51 trials and 8,952 participants analyzed, behaviour-change interventions produced the largest mean increase in daily steps (3,287 steps/day) versus usual care, but only multi-component interventions significantly improved every primary physical-activity measure (steps per day, moderate-to-vigorous activity, and combined activity) at once8. A time-course model embedded in the same analysis estimated that the benefit of these interventions peaked around 75 weeks post-enrollment, at roughly 738 additional steps per day above baseline, information that a simple before-after pairwise comparison at a single follow-up point could not have produced8. The secondary outcomes were more mixed: motivational/goal-setting and multi-component interventions produced small but statistically significant reductions in HbA1c (-0.28% and -0.24% respectively), while none of the four intervention designs produced a significant effect on weight, BMI, blood pressure or sedentary time, which indicates that an intervention shown to be effective for one outcome in a network cannot be assumed to be effective for every related outcome without separate supporting evidence8. The authors themselves reported high heterogeneity and low-to-moderate certainty across the network, and noted that most included trials followed participants only over the short term; both are standard caveats that a network meta-analysis inherits directly from the quality of the contributing trials rather than from the network modeling itself. Both case studies also illustrate a point relevant to how the findings of an exercise NMA are communicated to non-specialist readers: a ranked list of exercise types or intervention designs is easily presented as a simple “best to worst” table, but presenting it without the attached certainty ratings and confidence intervals risks implying a precision that the underlying evidence, by the authors’ own assessment, does not yet support. Presenting the certainty rating alongside the ranking, even in simplified form for a non-specialist audience, requires little space and preserves the transparency that the authors built into the analysis.

Methodological challenges of exercise trials for NMA

Both studies illustrate a challenge specific to exercise research: interventions labeled with the same name, such as “strength training” or “behaviour change”, can differ substantially in dose, frequency, supervision and adherence support across the pooled trials, which threatens the transitivity assumption on which NMA depends34. The PRISMA extension for network meta-analysis specifically requires authors to report how comparability across the network was assessed, which for exercise reviews typically means describing how intervention components, not just intervention labels, were coded and compared before pooling9. Both case studies also identified blinding as a near-universal limitation, since participants and staff almost always know whether an exercise intervention is being delivered; this structural source of risk of bias cannot be resolved by network geometry alone, and authors must carry it explicitly into their certainty ratings68. Researchers planning a new exercise-focused network should also allow time for the coding work this entails: building a shared taxonomy of intervention components across every included trial, with at least two reviewers independently coding each arm against it, is substantially more labor-intensive than extracting outcome data alone, but it is the step that determines whether the comparisons in the resulting network are valid. Reviewers of a submitted exercise NMA increasingly expect this taxonomy to be reported as a supplementary table rather than summarized in a single sentence of the methods section, since readers require the taxonomy itself to judge whether the transitivity assumption was reasonably met for the specific network, rather than merely asserted. Publishing this table alongside the main results also allows other research teams to reuse the same component taxonomy in a future update of the network, which reduces the cost of keeping an exercise evidence base current as new trials are published.

Key points

Key points

Network meta-analysis enables exercise and health-promotion researchers to rank specific intervention types against one another even in the absence of head-to-head trials. The value of the method, however, depends on coding intervention components carefully enough to satisfy transitivity and on carrying blinding-related risk of bias through into the certainty rating.

Practical application

Practical application

  1. The evidence base should be mapped as a network before an NMA is undertaken. Researchers should establish how many trials connect each pair of interventions indirectly through a shared comparator, and whether populations and designs are sufficiently similar to satisfy transitivity.
  2. Intervention components, not merely labels, should be coded during data extraction. Dose, frequency, supervision and adherence support should be recorded for each arm, since interventions sharing a name may differ sufficiently in these features to threaten the validity of the network.
  3. Certainty should be reported for each comparison using CINeMA or an equivalent framework. Exercise trials are rarely blinded, so certainty ratings should reflect this limitation explicitly rather than defaulting to a uniform rating across the network.

Frequently asked questions

No. NMA produces an estimate of how two interventions are likely to compare on the basis of indirect evidence, which is useful when a direct trial does not exist, but it is not a substitute for one, particularly when certainty ratings for the indirect comparison are low.

A large point estimate derived from only five comparisons and 107 participants carries a wide confidence interval and lower certainty, which is why the CINeMA framework rated most modalities other than walking or jogging as very low certainty rather than treating every ranked estimate as equally trustworthy.

Transitivity is the assumption that trials being connected through a shared comparator are similar enough in their populations, settings, and other effect modifiers that an indirect comparison between two interventions not studied head-to-head is still valid.

Ranking scores such as SUCRA (surface under the cumulative ranking curve) summarize the probability that one intervention is better than others across the network, but published guidance on interpreting NMA rankings warns that ranking probabilities should always be interpreted alongside the underlying effect sizes and their certainty, not as a standalone verdict.

In the Hodkinson et al. physical-activity NMA, multi-component interventions were the only category with significant effects across all primary measures, but the authors noted high heterogeneity and low-to-moderate certainty, so this finding should inform rather than determine the choice of intervention.

References

  1. Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Statistics in Medicine. 2004;23(20):3105-3124. read 10 October 2026. View on Wiley Online Library
  2. Caldwell DM, Ades AE, Higgins JPT. Simultaneous comparison of multiple treatments: combining direct and indirect evidence. BMJ. 2005;331(7521):897-900. read 10 October 2026. View on The BMJ
  3. Jansen JP, Fleurence R, Devine B, et al. Interpreting indirect treatment comparisons and network meta-analysis for health-care decision making: report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices: part 1. Value in Health. 2011;14(4):417-428. read 10 October 2026. View on Value in Health
  4. Salanti G. Indirect and mixed-treatment comparison, network, or multiple-treatments meta-analysis: many names, many benefits, many concerns for the next generation evidence synthesis tool. Research Synthesis Methods. 2012;3(2):80-97. read 10 October 2026. View on Research Synthesis Methods
  5. Jansen JP, Fleurence R, Devine B, et al. Interpreting indirect treatment comparisons and network meta-analysis for health-care decision making: report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices: part 1. Value in Health. 2011;14(4):417-428. read 10 October 2026. View on ISPOR
  6. Noetel M, Sanders T, Gallardo-Gómez D, et al. Effect of exercise for depression: systematic review and network meta-analysis of randomised controlled trials. BMJ. 2024;384:e075847. read 10 October 2026. View on The BMJ
  7. Nikolakopoulou A, Higgins JPT, Papakonstantinou T, et al. CINeMA: An approach for assessing confidence in the results of a network meta-analysis. PLOS Medicine. 2020;17(4):e1003082. read 10 October 2026. View on PLOS Medicine
  8. Hodkinson A, Sulochana HR, Kontopantelis E, et al. Effectiveness of different intervention designs for improving physical activity in adults with cardiometabolic conditions over time: a systematic review and network meta-analysis of randomised controlled trials. BMC Medicine. 2025;23:437. read 10 October 2026. View on BMC Medicine
  9. Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions. Annals of Internal Medicine. 2015;162(11):777-784. read 10 October 2026. View on Annals of Internal Medicine
  10. Mills EJ, Thorlund K, Ioannidis JPA. Demystifying trial networks and network meta-analysis. BMJ. 2013;346:f2914. read 10 October 2026. View on The BMJ

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 10). The Use of Network Meta-Analysis in Physical Exercise and Health Promotion Research. Pubrica. https://pubrica.com/academy/meta-analysis/the-use-of-network-meta-analysis-in-the-field-of-physical-exercise-and-health-promotion/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

Exercise NMA Feasibility Worksheet

These items should be checked before a network meta-analysis of exercise or health-promotion trials is undertaken.

[ ] Drawn a network diagram showing how every intervention connects through shared comparators
[ ] Coded intervention dose, frequency, supervision, and adherence support per arm, not just intervention labels
[ ] Checked whether populations and settings across connected trials are similar enough for transitivity
[ ] Selected a certainty-rating framework (e.g. CINeMA) and planned how blinding limitations will be reflected in it
[ ] Confirmed the relevant PRISMA-NMA items will be reported alongside base PRISMA 2020

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica’s evidence synthesis team conducts exercise and health-promotion network meta-analyses from start to finish, from network feasibility mapping and coding of intervention components to PRISMA-NMA-compliant reporting and CINeMA certainty ratings.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Feasibility of a network meta-analysis for an exercise evidence base

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.